AI-GENERATED DEEPFAKES AND THE CONSTITUTIONAL BOUNDARIES OF CRIMINAL LIABILITY: A DOCTRINAL EXAMINATION OF THE PRINCIPLE OF LEGALITY, CRIMINAL INTENT, AND MEXICO’S QUINTANA ROO PENAL CODE CONTROVERSY

Abstract

The proliferation of artificial intelligence-generated deepfakes has complicated traditional approaches to criminal responsibility by enabling convincing synthetic representations capable of facilitating deception, impersonation, reputational injury, and other unlawful conduct. This study examines the constitutional boundaries of criminal liability for deepfake-related wrongdoing through a doctrinal analysis of the principle of legality, criminal intent, and the controversy surrounding Article 20 Bis of Mexico’s Quintana Roo Penal Code. Introduced through Decree 211 in April 2024, Article 20 Bis authorizes an increase of up to one-half in the prescribed penalty when artificial intelligence is used to commit an existing criminal offense. The provision was challenged in Action of Unconstitutionality 104/2024 on grounds including legal certainty and statutory precision. Using qualitative doctrinal methodology, the research analyzes statutory language, constitutional safeguards, judicial reasoning, and established concepts of culpability, while applying the resulting framework to hypothetical deepfake scenarios. Particular attention is given to foreseeability, the prohibition of analogical expansion of criminal offenses, the relationship between technological means and prohibited conduct, and the evidentiary requirements for establishing intent. The analysis identifies three interrelated difficulties: the breadth of statutory definitions of artificial intelligence, the risk of treating technological involvement as a substitute for proof of individual culpability, and uncertainty when synthetic content crosses the boundary between protected expression and punishable harm. In April 2025, Mexico’s Supreme Court dismissed the constitutional action after a proposal to invalidate the provision failed to obtain the required qualified majority; this procedural outcome should not be confused with a definitive judicial endorsement of every aspect of its wording. The study argues that constitutionally defensible deepfake enforcement requires clearly specified underlying offenses, individualized assessment of knowledge and intent, demonstrable links between conduct and legally relevant harm, and proportionate sentencing. It proposes technology-neutral statutory drafting complemented by narrowly tailored AI-related aggravating factors and safeguards for legitimate expression. These findings contribute a structured approach to reconciling responses to synthetic-media misuse with legality, due process, and fundamental rights within Mexican criminal law and wider, internationally relevant comparative debates.